tokenize-text

Javascript text tokenizer that is easy to use and compose.

Installation

$ npm install tokenize-text

Usage

var Tokenizer = require('tokenize-text');
var tokenize = new Tokenizer();

tokenize.split(fn)

This is the main method of this module, all other methods are using it.

fn will be called with 4 arguments:

• text: text value of the token (text == currentToken.value)
• currentToken: current token object
• prevToken: precedent token (or null)
• nextToken: next token (or null)

fn should return a string, an array of string, a token or an array of tokens.

tokenize.split(fn) returns a tokenizer function that accept a list of tokens or a string argument (it will be convert as one token).

The tokenizer function returns an array of tokens with the following properties:

• value: text content of the token
• index: absolute position in the original text
• offset: length of the token (equivalent to value.length)

// Simple tokenizer that split into 2 sections
var splitIn2 = tokenize.split(function(text, currentToken, prevToken, nextToken) {
    return [
        text.slice(0, text.length / 2),
        text.slice(text.length / 2)
    ]
});

var tokens = splitIn2('hello');

/*
[
    { value: 'he', index: 0, offset: 2 },
    { value: 'llo', index: 2, offset: 3 }
]
*/

tokenize.re(re)

Tokenize using a regular expression:

var extractUppercase = tokenize.re(/[A-Z]/);
var tokens = extractUppercase('aBcD');

/*
[
    { value: 'B', index: 1, offset: 1 },
    { value: 'D', index: 3, offset: 1 }
]
*/

tokenize.characters()

Tokenize and split as characters, tokenize.characters() is equivalent to tokenize.re(/[^\s]/).

var tokens = tokenize.characters()('abc');

/*
[
    { value: 'a', index: 0, offset: 1 },
    { value: 'b', index: 1, offset: 1 },
    { value: 'c', index: 2, offset: 1 }
]
*/

tokenize.sections()

Split in sections, sections are split by \n . , ; ! ?.

var tokens = tokenize.sections()('this is sentence 1. this is sentence 2');

/*
[
    {
        value: 'this is sentence 1',
        index: 0,
        offset: 18
    },
    {
        value: ' this is sentence 2',
        index: 19,
        offset: 19
    }
]
*/

tokenize.words()

Split in words:

var tokens = tokenize.words()('hello, how are you?');

/*
[
    { value: 'hello', index: 0, offset: 5 },
    { value: 'how', index: 7, offset: 3 },
    { value: 'are', index: 11, offset: 3 },
    { value: 'you', index: 15, offset: 3 }
]
*/

tokenize.filter(fn)

Filter the list of tokens by calling fn(token):

// Filter the words to extract the ones that start with an uppercase
var extractNames = tokenize.filter(function(word, current, prev) {
    return (prev && /[A-Z]/.test(word[0]));
});

// Split texts in words
var words = tokenize.words()('My name is Samy.');

// Apply the filter
var tokens = extractNames(words);

/*
[
    { value: 'Samy', index: 11, offset: 4 }
]
*/

tokenize.flow(fn1, fn2, [...])

Creates a tokenizer that returns the result of invoking the provided tokenizers for each input token.

var extractNames = tokenize.flow(
    // Split text as words
    tokenize.words(),

    // Filter the words to extract the ones that start with an uppercase
    tokenize.filter(function(word, current, prev) {
        return (prev && /[A-Z]/.test(word[0]));
    })
);

var tokens = extractNames('My name is Samy.');

To execute all tokenizer in series, you can use tokenize.serie(fn1, fn2, [...]) instead.

Examples

Extract repeated words in sentences

Example to extract all repeated words in sentences:

var repeatedWords = tokenize.flow(
    // Tokenize as sections
    tokenize.sections(),

    // For each sentence
    tokenize.flow(
        // Tokenize as words
        tokenize.words(),

        // Filter words to extract only repeated ones
        tokenize.filter(function(word, token, prev) {
            return (
                prev &&
                token.value.toLowerCase() === prev.value.toLowerCase()
            );
        })
    )
);


var tokens = repeatedWords('This is great great. Great is an an awesome words');

/*
[
    { value: 'great', index: 14, offset: 5 },
    { value: 'an', index: 33, offset: 2 }
]
*/